For the video version, click here.

For most people, negative trials are disappointing – they are a letdown. We love the flashy headlines when a new therapy seems to treat something that was previously untreatable. But negative trials are arguably the most important part of medical science. They drive the field in new directions. They blow up superstitions and cultural practices and standards of care that are based on nothing more than expert opinion. I love negative trials.

But this one was a bit of a letdown.

Appearing in the Journal of the American Medical Association, we have a randomized trial evaluating music therapy for children with autism spectrum disorder. I really wanted this one to work.

This was a multicenter trial which randomized 364 kids with ASD to standard care versus improvisational music therapy. In the latter arm, trained music therapists would work with the children from 1 to 3 times a week for five months, following their lead and improvising with them using techniques like synchronization and mirroring. This feels right – here is a syndrome characterized by deficits in social communication – perhaps engagement through music would help.

But it didn't. Children in the music group had no greater improvement in social affect than those in the control group out to a year of follow-up, though both groups improved a bit.

When a study is negative, there is one important question to ask? Was it the intervention that didn't work, or was it the study that didn't work?

And this study had some problems. I'll highlight the two major ones.

First, the authors examined only individuals that attended at least one follow-up after randomization.

14% of those randomized were never analyzed, and the standard care group had more of those kids – you can see the discrepancy between the orange and blue bars in this chart.

This could bias the results against music therapy if more severely affected kids were more likely to withdraw after being randomized into the usual care arm.

Second, the study was stopped early due, it seems, to lack of funding, leaving only 42% power to detect the clinically meaningful difference between the arms they were looking for.

I should also mention that interventions like this are hard to assess because they aren't uniform. It's not a drug that has the same composition and dosage for each patient. Music therapy would differ from patient to patient and from therapist to therapist. This non-precision of exposure tends to bias results towards the null hypothesis.

So what we have here is a negative trial, which under ordinary circumstances would steer us away from music therapy as a viable autism intervention, but with some major caveats. If this is to be tested again, we'll need a larger trial, with more robust follow-up, more uniform therapeutic practices, and, frankly, more money for recruitment.